CN105955710A - Parallel simulation data processing method based on tree communication structure - Google Patents
Parallel simulation data processing method based on tree communication structure Download PDFInfo
- Publication number
- CN105955710A CN105955710A CN201610255471.6A CN201610255471A CN105955710A CN 105955710 A CN105955710 A CN 105955710A CN 201610255471 A CN201610255471 A CN 201610255471A CN 105955710 A CN105955710 A CN 105955710A
- Authority
- CN
- China
- Prior art keywords
- node
- calculating
- communication
- layer
- tree
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/30—Arrangements for executing machine instructions, e.g. instruction decode
- G06F9/38—Concurrent instruction execution, e.g. pipeline, look ahead
- G06F9/3818—Decoding for concurrent execution
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/44—Arrangements for executing specific programs
- G06F9/455—Emulation; Interpretation; Software simulation, e.g. virtualisation or emulation of application or operating system execution engines
Abstract
The invention discloses a parallel simulation data processing method based on a tree communication structure. The method comprises: setting a parallel algorithm on a selected simulation model; reading data parameters of the simulation model, the data parameters being prestored in a database; according to the data parameters, determining subsystems and the number N of systems in a coupling simulation system; according to load balancing of each computational node, distributing the N subsystems to each computational node; determining a tree communication link connection structure of each computational node; determining an information flow method and an information exchange structure between each computational node in the tree communication link connection structure; and each computational node executing calculation and outputting a final computation result. Using the method, parallel simulation computation speed of high-coupling computing objects in a high-performance computer can be improved, so as to reduce communication cost and realize a parallel simulation calculation task on a complex big data simulation system.
Description
Technical field
The present invention relates to field of computer technology, particularly relate at a kind of parallel artificial data based on tree-like communication structure
Reason method.
Background technology
During the calculating of modern supercomputer, when calculate to as if during the complication system of a high coupling, often need
To calculate node more and carry out parallel computation to obtain final calculation result.But, the parallel computation of supercomputer is calculating effect
Having reached bottleneck in rate, the practicality causing supercomputer is relatively low.This bottleneck is the meter in each execution calculating task
Under the amount of calculation non-load balanced case of operator node, increase calculating node and can shorten the calculating time, but be as calculating node
Increase and also result in the different communication time calculated between node prolongation.Generally, consume along with the increase calculating node
The calculating time of expense is linear minimizing, along with the increase calculating node causes the communication time calculated between node then to refer to
Several levels increases.Therefore, calculate node quantity increase to a certain amount of after, whole computer parallel computation will necessarily occur
The situation that time reduces.This is not only the key issue affecting computational efficiency raising run into during computer parallelization calculates
One of, also it is that modern computer parallelization calculating field is difficult to write the general commercial software being applicable to high coupling calculating system
One of difficult point.
Summary of the invention
The embodiment of the present invention proposes a kind of parallel artificial data processing method based on tree-like communication structure, improves in high property
Computer can carry out the high parallel artificial calculating speed coupling and calculating object, reduce communication-cost, it is achieved to complicated number greatly
The task of parallel artificial calculating is carried out according to analogue system.
The embodiment of the present invention provides a kind of parallel artificial data processing method based on tree-like communication structure, including:
Selected phantom is set parallel algorithm;
Read the data parameters that described phantom is pre-stored in data base;
According to described data parameters, determine the subsystem in coupling Simulation system and system number N;
According to the load balancing of each calculating node, N number of subsystem is distributed to each described calculating node;
Determine the tree-like communication link attachment structure between each calculating node;Wherein, described tree-like communication link connects knot
Structure is formed tree-like attachment structure by the communication link network between M described calculating node;Each layer in described communication link network
The quantity calculating node successively decrease successively, and the arbitrary calculating node in i-th layer is only sent out with a calculating node in i+1 layer
Raw communication, there is not communication between node in each calculating theing be in same layer;I >=2, M >=3;
Determine each information flow mode calculated between node and information exchange in described tree-like communication link attachment structure
Structure;Wherein, described information switching fabric be each level calculate node only exchange through information category filter after calculating knot
Really;
Each node that calculates performs calculate and export final calculation result.
Further, described communication link network is to arrange to meet following condition:
(1) k is setijIt is i-th layer of jth calculating node in described communication link network, i=1,2 ..., r, j=1,2 ...,
s;
(2) K is setiα={ x x represents several k of i-th layerijAn element in the α set that element is constituted };
(3) K is setiIt is i-th layer of whole calculating node kijThe nonvoid set constituted;
If KiSome separate subset Kiα, α ∈ I, then meet following condition:
①Kiα, α ∈ I mutually disjoints two-by-two, i.e. as α ≠ β, always has
②Kiα, α ∈ be I's and be Ki, i.e. ∪α∈IKiα=Ki;
Wherein, KiαRepresent several calculating node k of i-th layerijThe α the set constituted for element;KiRepresent by subset
Close KiαThe nothing constituted is handed over and family of subsets, α=1, and 2 ..., n, niRepresent the subclass K of i-th layerizNumber.
Further, the arbitrary calculating node in described i-th layer only occurs logical with a calculating node in i+1 layer
News, particularly as follows:
Arbitrary calculating node k in described i-th layeriOnly with the calculating node k in i+1 layeri+1zThere is communication.
Further, described according to described data parameters, determine the subsystem in coupling Simulation system and system number N,
Particularly as follows:
According to analyzing the essential information of object, topological connection relation and the control parameter of running, determine coupling Simulation
Subsystem in system and system number N.
Further, the described load balancing according to each calculating node, N number of subsystem is distributed to each described calculating and saves
Point, particularly as follows:
According to calculate the number of node, the number of CPU core of each calculating node, each calculating node memory size with
And artificial tasks number, premised on load balancing, N number of subsystem is distributed to each described calculating node.
Implement the embodiment of the present invention, have the advantages that
The parallel artificial data processing method based on tree-like communication structure that the embodiment of the present invention provides, to selected emulation
Model specification parallel algorithm, read corresponding data parameters and determine the coupling Simulation heart quiet in subsystem and system number.So
Afterwards according to the load balancing of each calculating node, subsystem is distributed to each calculating node, then determines the tree between each calculating node
Shape communication link attachment structure.The different communications calculated between node that the present invention utilizes in tree topology in each layer are even
Contact constantly reduces, and the feature that corresponding communication link is necessarily greatly reduced when from bottom to top going forward one by one, and determines each meter
Communication link between operator node, it is ensured that in every layer, arbitrary calculating node only communicates with a calculating node of last layer,
And be between each calculating node of same layer and do not communicate.Furthermore, determine each calculating node information flow mode and
Information switching fabric, when the calculating node that information switching fabric is each level here sends communication, only exchanges through information kind
Result of calculation after class filtration.Last each node that calculates performs calculate and export final calculation result.Compared to prior art meeting
Running into the bottleneck of parallel efficiency calculation, the present invention constructs a tree-like communication link attachment structure of overall importance, it is ensured that each layer
The result of calculation uniqueness that level is uploaded, it is to avoid cause too much communication-cost owing to calculating the increase of node, improves parallel meter
The computational efficiency calculated.It addition, when each calculating node carries out information exchange, each node that calculates only exchanges the knot of independent calculating
Really, filtered the program of exchange information by structure hierarchical grouping, during reducing the parallel computation of the overall situation, need the information of communication
Kind.Finally, the information of communication between node that calculates each different layers successively specifies transfer mode, limits communication information
Some transfer mode on different layers, reduces the semaphore of each communication information in communication process.
Accompanying drawing explanation
Fig. 1 is a kind of embodiment of the parallel artificial data processing method based on tree-like communication structure that the present invention provides
Schematic flow sheet;
Fig. 2 is the structural representation of a kind of embodiment of the tree-like communication structure that the present invention provides.
Detailed description of the invention
Below in conjunction with the accompanying drawing in the embodiment of the present invention, the technical scheme in the embodiment of the present invention is carried out clear, complete
Describe, it is clear that described embodiment is only a part of embodiment of the present invention rather than whole embodiments wholely.Based on
Embodiment in the present invention, it is every other that those of ordinary skill in the art are obtained under not making creative work premise
Embodiment, broadly falls into the scope of protection of the invention.
See Fig. 1, be that the one of the parallel artificial data processing method based on tree-like communication structure that the present invention provides is real
Executing the schematic flow sheet of example, the method comprising the steps of 101 to 107, and this step is specific as follows:
Step 101: selected phantom is set parallel algorithm.
Step 102: read the data parameters that phantom is pre-stored in data base.
Step 103: according to data parameters, determines the subsystem in coupling Simulation system and system number N.
In the present embodiment, step 103 is particularly as follows: according to analyzing the essential information of object, topological connection relation and operation
The control parameter of process, determines the subsystem in coupling Simulation system and system number N.
Step 104: according to the load balancing of each calculating node, distributes to each calculating node by N number of subsystem.
In the present embodiment, step 104 particularly as follows: according to calculate the number of node, each calculating node CPU core
The memory size of calculating node several, each and artificial tasks number, premised on load balancing, distribute to N number of subsystem
Each described calculating node.
Step 105: determine the tree-like communication link attachment structure between each calculating node;Wherein, tree-like communication link is even
Access node structure is formed tree-like attachment structure by M the communication link network calculated between node;The meter of each layer in communication link network
The quantity of operator node is successively decreased successively, and the arbitrary calculating node in i-th layer only occurs logical with a calculating node in i+1 layer
News, there is not communication between node in each calculating theing be in same layer;I >=2, M >=3.
Step 106: determine each information flow mode calculated between node and information in tree-like communication link attachment structure
Switching fabric;Wherein, information switching fabric be each level calculate node only exchange through information category filter after calculating knot
Really.
Step 107: each node that calculates performs calculate and export final calculation result.
In the present embodiment, communication link network is to arrange to meet following condition:
(1) k is setijIt is i-th layer of jth calculating node in described communication link network, i=1,2 ..., r, j=1,2 ...,
s;
(2) K is setiα={ x x represents several k of i-th layerijAn element in the α set that element is constituted };
(3) K is setiIt is i-th layer of whole calculating node kijThe nonvoid set constituted;
If KiSome separate subset Kiα, α ∈ I, then meet following condition:
①Kiα, α ∈ I mutually disjoints two-by-two, i.e. as α ≠ β, always has
②Kiα, α ∈ be I's and be Ki, i.e. ∪α∈IKiα=Ki;
Wherein, KiαRepresent several calculating node k of i-th layerijThe α the set constituted for element;KiRepresent by subset
Close KiαThe nothing constituted is handed over and family of subsets, α=1, and 2 ..., n, niRepresent the subclass K of i-th layerizNumber.
In the present embodiment, when the leaf layer of tree structure is referred to as the ground floor at the end by us when, ground floor complete
Portion calculates node can be divided into some subclass Kiz, any one subclass of ground floor calculating with last layer all only
Node ki+1zCommunication relation occurs, and there is not communication relation in the subset of same level, thus constitutes one uniquely to calculate node
ki+1zLocal area communication net for the single switch on summit.And then, other subclass of ground floor are constituted the most in the same fashion
The local area communication net of different single switches.Meanwhile, the local area communication net of whole single switches constitutes the base of ground floor
One layer of tree-like communication network combination in the local area communication net of single switch.Finally, we can construct the tree-like of last layer
Communication network combines.And so on, we can finally constitute the local area communication net of the most unique top layer switch.As
Shown in Fig. 2, Fig. 2 is the structural representation of a kind of embodiment of the tree-like communication structure that the present invention provides.
In the present embodiment, the arbitrary calculating node in i-th layer only occurs logical with a calculating node in i+1 layer
News, particularly as follows: the arbitrary calculating node k in i-th layeriOnly with the calculating node k in i+1 layeri+1zThere is communication.
In the present embodiment, the present invention except structure tree structure communication link network in addition to, also to global calculation during
The different information categories calculated between nodes carry out filtering reduction so that the different differences calculating nodes can pass through a mark
Amount, one group of scalar or other less measurement units represent.So make to calculate the standardisation process that node represents, Ke Yishi
During now calculating, the different transinformations calculating node tail off.The most just can be by transmitting less signal,
Reach to transmit the effect of bulk information.
In the present embodiment, the present invention except structure tree structure communication link, carry out the filtration of information category in addition to, go back structure
Make communication signal transmissions structure of overall importance.This structure can reduce the transmission semaphore of communication process, improves parallel computation
Efficiency.
Technical scheme can realize three below effect: 1, reduces communication link;2, the kind of communication information is reduced
Class;3 signals reducing communication.In order to better illustrate the effect that technical solution of the present invention can realize, it is respectively directed to above-mentioned technology
Effect elaborates.
One, communication link is reduced
In the ordinary course of things, during the simulation object to Global-Coupling can specify that single layer structure communication link network structure
All all calculating node mutual communication link networks between any two that communication link is only a layer calculated between node,
Thus, the present invention can set up corresponding communication link network of overall importance, referred to as a communication link network for single layer structure.
In the communication network of overall importance of above this type communication structure definition, it is assumed that calculate between node for every a pair
An one-way communication link be 1, the round both-way communication link that every a pair calculates between node is 2, then tie at monolayer
Total amount p of whole communication links required for all calculating calculating processes whole between node in structure communication link network structure1
It is equal to:
p1=ni(ni1)=ni 2-ni
And the ground floor at the end of the tree structure communication network combination that the present invention builds has niIndividual subset, therefore, each
Whole calculating nodes that height is concentrated can generate at least 1, at mostIndividual and upper strata calculates and there is mutual communication between node
The communication network combination that the local area communication net of the single switch of relation is constituted.Obviously, each subset of ground floor is whole
The total amount calculating the communication link that node calculates node with upper strata is 2ni.And so on, in tree structure communication network combines
Carry out once calculating the whole communication link p required for the calculating process of overall importance that node is relevant to each2It is equal to:
Owing to the second communication structure defined above must haveRelation, therefore obtain conclusion:
(1) when number n calculating node of i-th layeriDuring more than 5, the local area communication net of the single switch of any i+1 layer
Between the total amount of communication link necessarily reduce with the speed of each multiple determined, this multiple is equal to i-th layer
Calculating the number business divided by the number calculating node of i+1 layer of node, meanwhile, this business must be greater than or equal to 2
One real number;
(2) when number n calculating node of i-th layeriDuring more than 5, tree structure communication network combination carry out once with often
One whole communication link calculated required for the calculating process of overall importance that node is all correlated with must be less than the first type
Communication link.
(3) when number n calculating node of i-th layeriMore than 5, and number n calculating node of i-th layeriNumerical value the biggest
When, the combination of tree structure communication network carries out once calculating the calculating process institute of overall importance that node is relevant to each
The whole communication links needed must be the biggest less than the difference of the communication link of the first type.
Two, the kind of communication information is decreased
During the minimizing of communication information kind can greatly reduce Global-Coupling parallel computation, different calculating nodes it
Between the information category of transmission, thus reach to improve the situation that tree-like communication structure root information communication is blocked up.Specifically, in tree
The ground floor of shape fabric communication combination of network has n1Individual subset, therefore can generate at least 1, at mostIndividual set with upper strata deposits
In the communication network combination that the local area communication net of the single switch of mutual communication relation is constituted, then ground floor subset and upper strata collection
The communication kind closed is 2n1u.And so on, carry out once calculating node with each in tree structure communication network combines
Required whole communication kinds U realized of the calculating process of overall importance being all correlated with2It is equal to:
Owing to the second communication structure defined above must haveRelation, therefore obtain conclusion:
(1) when number n calculating node of i-th layeriDuring more than 5, the local area communication net of the single switch of any i+1 layer
Communication information kind and necessarily reduce with the speed of each multiple determined, this multiple is equal to i-th layer
Calculating the number business divided by the number calculating node of i+1 layer of node, meanwhile, this business must be greater than or equal to 2
One real number;
(2) when number n calculating node of i-th layeriDuring more than 5, tree structure communication network combination carry out once with often
The kind of one whole communication information calculating the required realization of the calculatings process of overall importance all be correlated with of node must be less than the
The kind of a type of communication information.
Three, the signal of communication is decreased
The known ground floor in the combination of tree structure communication network has n1Individual subset, therefore can generate at least 1, at mostIt is individual that there is the communication network combination that the local area communication net of the single switch of mutual communication relation is constituted in set with upper strata, then and the
One straton integrates the communication kind with upper strata set as 2n1u.Given that it is known that carry out the behaviour of primary information issue and collection in overall situation tree
Making to need to calculate node contacts with each, the kind carrying out communication between each the calculating node on the most each layer is carried out
The semaphore of the communication required for communication is known constant m, then, in tree structure communication network combines, carry out one
Secondary required whole communication signal amounts S realized of calculating process of overall importance calculating node to each relevant are equal to:
Visible, it is transmitted, at this transmission side communication by the information category filtered during overall situation parallel computation
Under formula, carrying out the required whole communication signal amounts realized of calculating process the most of overall importance can be greatly reduced.
In sum, the parallel artificial data processing method based on tree-like communication structure that the embodiment of the present invention provides is right
Selected phantom sets parallel algorithm, read corresponding data parameters and determine the coupling Simulation heart quiet in subsystem and be
System number.Then according to the load balancing of each calculating node, subsystem is distributed to each calculating node, then determines each calculating node
Between tree-like communication link attachment structure.Between the different calculating nodes that the present invention utilizes in tree topology in each layer
Communication junction point constantly reduce, and the feature that corresponding communication link is necessarily greatly reduced when from bottom to top going forward one by one,
Determine the communication link between each calculating node, it is ensured that in every layer, arbitrary calculating node is sent out with a calculating node of last layer
Raw communication, and be between each calculating node of same layer and do not communicate.Furthermore, determine the information flow of each calculating node
Mode and information switching fabric, when the calculating node that information switching fabric is each level here sends communication, only exchange process
Result of calculation after information category filtration.Last each node that calculates performs calculate and export final calculation result.Compared to existing
Technology can run into the bottleneck of parallel efficiency calculation, and the present invention constructs a tree-like communication link attachment structure of overall importance, protects
Demonstrate,prove the result of calculation uniqueness that each level is uploaded, it is to avoid cause too much communication-cost owing to calculating the increase of node, improve
The computational efficiency of parallel computation.It addition, when each calculating node carries out information exchange, each node that calculates only exchanges independent calculating
Result, by structure hierarchical grouping filter exchange information program, reduce the overall situation parallel computation during need communication
Information category.Finally, the information of communication between node that calculates each different layers successively specifies transfer mode, limits communication
Some transfer mode on information different layers, reduces the semaphore of each communication information in communication process.
One of ordinary skill in the art will appreciate that all or part of flow process realizing in above-described embodiment method, be permissible
Instructing relevant hardware by computer program to complete, described program can be stored in a computer read/write memory medium
In, this program is upon execution, it may include such as the flow process of the embodiment of above-mentioned each method.Wherein, described storage medium can be magnetic
Dish, CD, read-only store-memory body (Read-Only Memory, ROM) or random store-memory body (Random Access
Memory, RAM) etc..
The above is the preferred embodiment of the present invention, it is noted that for those skilled in the art
For, under the premise without departing from the principles of the invention, it is also possible to make some improvements and modifications, these improvements and modifications are also considered as
Protection scope of the present invention.
Claims (5)
1. a parallel artificial data processing method based on tree-like communication structure, it is characterised in that including:
Selected phantom is set parallel algorithm;
Read the data parameters that described phantom is pre-stored in data base;
According to described data parameters, determine the subsystem in coupling Simulation system and system number N;
According to the load balancing of each calculating node, N number of subsystem is distributed to each described calculating node;
Determine the tree-like communication link attachment structure between each calculating node;Wherein, described tree-like communication link attachment structure is by M
Communication link network between individual described calculating node forms tree-like attachment structure;The calculating of each layer in described communication link network
The quantity of node is successively decreased successively, and the arbitrary calculating node in i-th layer only occurs logical with a calculating node in i+1 layer
News, there is not communication between node in each calculating theing be in same layer;I >=2, M >=3;
Determine each information flow mode calculated between node and information switching fabric in described tree-like communication link attachment structure;
Wherein, described information switching fabric be each level calculate node only exchange through information category filter after result of calculation;
Each node that calculates performs calculate and export final calculation result.
Parallel artificial data processing method based on tree-like communication structure the most according to claim 1, it is characterised in that institute
Stating communication link network is to arrange to meet following condition:
(1) k is setijIt is i-th layer of jth calculating node in described communication link network, i=1,2 ..., r, j=1,2 ..., s;
(2) K is setiα={ x x represents several k of i-th layerijAn element in the α set that element is constituted };
(3) K is setiIt is i-th layer of whole calculating node kijThe nonvoid set constituted;
If KiSome separate subset Kiα, α ∈ I, then meet following condition:
①Kiα, α ∈ I mutually disjoints two-by-two, i.e. as α ≠ β, always has
②Kiα, α ∈ be I's and be Ki, i.e. ∪α∈IKiα=Ki;
Wherein, KiαRepresent several calculating node k of i-th layerijThe α the set constituted for element;KiRepresent by subclass Kiα
The nothing constituted is handed over and family of subsets, α=1, and 2 ..., n, niRepresent the subclass K of i-th layerizNumber.
Parallel artificial data processing method based on tree-like communication structure the most according to claim 2, it is characterised in that institute
State the arbitrary calculating node in i-th layer and only calculate node generation communication with in i+1 layer, particularly as follows:
Arbitrary calculating node k in described i-th layeriOnly with the calculating node k in i+1 layeri+1zThere is communication.
Parallel artificial data processing method based on tree-like communication structure the most according to claim 1, it is characterised in that institute
State according to described data parameters, determine the subsystem in coupling Simulation system and system number N, particularly as follows:
According to analyzing the essential information of object, topological connection relation and the control parameter of running, determine coupling Simulation system
In subsystem and system number N.
Parallel artificial data processing method based on tree-like communication structure the most according to claim 1, it is characterised in that institute
State the load balancing according to each calculating node, N number of subsystem is distributed to each described calculating node, particularly as follows:
According to calculating the number of node, the number of CPU core of each calculating node, the memory size of each calculating node and imitating
True task number, premised on load balancing, distributes to each described calculating node by N number of subsystem.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610255471.6A CN105955710B (en) | 2016-04-22 | 2016-04-22 | Parallel artificial data processing method based on tree-like communication structure |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610255471.6A CN105955710B (en) | 2016-04-22 | 2016-04-22 | Parallel artificial data processing method based on tree-like communication structure |
Publications (2)
Publication Number | Publication Date |
---|---|
CN105955710A true CN105955710A (en) | 2016-09-21 |
CN105955710B CN105955710B (en) | 2019-03-01 |
Family
ID=56915155
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610255471.6A Active CN105955710B (en) | 2016-04-22 | 2016-04-22 | Parallel artificial data processing method based on tree-like communication structure |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105955710B (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108595211A (en) * | 2018-01-05 | 2018-09-28 | 百度在线网络技术(北京)有限公司 | Method and apparatus for output data |
CN112445675A (en) * | 2019-09-02 | 2021-03-05 | 无锡江南计算技术研究所 | Large-scale parallel program performance data rapid collection method based on layer tree network |
CN113556242A (en) * | 2020-04-24 | 2021-10-26 | 中科寒武纪科技股份有限公司 | Method and equipment for performing inter-node communication based on multi-processing nodes |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101311917A (en) * | 2007-05-24 | 2008-11-26 | 中国科学院过程工程研究所 | Particle model faced multi-tier direct-connection cluster paralleling computing system |
CN101883039A (en) * | 2010-05-13 | 2010-11-10 | 北京航空航天大学 | Data transmission network of large-scale clustering system and construction method thereof |
CN104702690A (en) * | 2015-03-12 | 2015-06-10 | 杭州域竹科技有限公司 | Distributed high-performance computing method based on virtual tree network technology |
-
2016
- 2016-04-22 CN CN201610255471.6A patent/CN105955710B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101311917A (en) * | 2007-05-24 | 2008-11-26 | 中国科学院过程工程研究所 | Particle model faced multi-tier direct-connection cluster paralleling computing system |
CN101883039A (en) * | 2010-05-13 | 2010-11-10 | 北京航空航天大学 | Data transmission network of large-scale clustering system and construction method thereof |
CN104702690A (en) * | 2015-03-12 | 2015-06-10 | 杭州域竹科技有限公司 | Distributed high-performance computing method based on virtual tree network technology |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108595211A (en) * | 2018-01-05 | 2018-09-28 | 百度在线网络技术(北京)有限公司 | Method and apparatus for output data |
CN108595211B (en) * | 2018-01-05 | 2021-11-26 | 百度在线网络技术(北京)有限公司 | Method and apparatus for outputting data |
CN112445675A (en) * | 2019-09-02 | 2021-03-05 | 无锡江南计算技术研究所 | Large-scale parallel program performance data rapid collection method based on layer tree network |
CN112445675B (en) * | 2019-09-02 | 2022-09-13 | 无锡江南计算技术研究所 | Large-scale parallel program performance data rapid collection method based on layer tree network |
CN113556242A (en) * | 2020-04-24 | 2021-10-26 | 中科寒武纪科技股份有限公司 | Method and equipment for performing inter-node communication based on multi-processing nodes |
CN113556242B (en) * | 2020-04-24 | 2023-01-17 | 中科寒武纪科技股份有限公司 | Method and equipment for performing inter-node communication based on multi-processing nodes |
Also Published As
Publication number | Publication date |
---|---|
CN105955710B (en) | 2019-03-01 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105049353B (en) | A kind of method and controller for business configuration routed path | |
CN105469143B (en) | Network-on-chip method for mapping resource based on neural network dynamic feature | |
CN108399464A (en) | A kind of multimodal transport method for optimizing route and system | |
CN104036324B (en) | Optimal design method of communication network capacity based on genetic algorithm | |
CN104185242B (en) | A kind of wireless Mesh netword channel distribution for reducing interference and route combined optimization system and method | |
CN106936645B (en) | The optimization method of tree network topology structure based on queueing theory | |
CN107222348A (en) | A kind of method for reducing power information physical system cascading failure risk | |
CN105279615A (en) | Active power distribution network frame planning method on the basis of bi-level planning | |
CN105976048A (en) | Power transmission network extension planning method based on improved artificial bee colony algorithm | |
CN103298009B (en) | The modeling of complicated wireless network control system and fault propagation method | |
CN105955710A (en) | Parallel simulation data processing method based on tree communication structure | |
CN103036792A (en) | Transmitting and scheduling method for maximizing minimal equity multiple data streams | |
CN105429185A (en) | Economic dispatching method with robust collaborative consistency | |
CN103384354A (en) | Optimum design method of optical distribution network of passive optical network | |
CN102904815A (en) | Scale-free network based router-level topology modeling method | |
CN104618130A (en) | Minimum cost synchronization method of data center network controller by using software | |
CN106685745A (en) | Network topology construction method and device | |
CN114727353A (en) | Self-adaptive information transmission method of mobile opportunistic sensor network based on genetic algorithm | |
CN109889525A (en) | Multi-communication protocol Intellisense method | |
CN108711860A (en) | A kind of power distribution network substation-lines join planing method based on parallel computation | |
CN104410529A (en) | Minimum cover deployment method for controllers based on software definition data center network | |
CN106789641A (en) | The topology design method of centralized wide area protection communication network in intelligent grid | |
CN109033603B (en) | Intelligent substation secondary system simulation method based on source flow path chain | |
CN112001141B (en) | Brain network inspired middle-large scale on-die interconnection system comprehensive method | |
CN106550387B (en) | A kind of wireless sensor network routing layer QoS evaluating method |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |